What FDA’s Recent Rare Disease Approvals Teach Us About Single‑Arm Trial Design

Between late 2024 and late 2025, FDA approved six rare-disease therapies supported primarily by single-arm trials. None of these sponsors ran randomized controlled trials. All received traditional or accelerated approval.

What separated success from rejection wasn’t luck or regulatory leniency. It was understanding what evidence compensates for the absence of randomization.

Single-arm trials are often positioned as a last resort, reserved for settings with tiny populations, severe diseases, and unforgiving timelines. Recent FDA approvals tell a more precise story. Single-arm designs are not inherently weak; they succeed when embedded in an evidence strategy that makes causal interpretation unavoidable.

Case Study 1: Papzimeos (Recurrent Respiratory Papillomatosis)

Design

Single-arm, open-label Phase 1/2 dose-escalation with expansion.

Population context

Approximately 27,000 adults in the U.S. live with recurrent respiratory papillomatosis (RRP). The disease was rare but not ultra-rare; critically, no FDA-approved therapies existed at the time of development.

Clinical baseline

Patients with RRP undergo frequent surgical debulking to maintain airway patency. Surgery frequency before enrollment provided a clear, quantifiable baseline for disease burden.

Why the single-arm comparison worked

The disease course of RRP is well understood and relatively stable absent intervention, making within-patient pre- vs post-treatment comparisons interpretable rather than speculative. The endpoint was not a soft symptom score or surrogate; it was a reduction in surgeries, a clinical outcome tightly coupled to patient burden.

In the pivotal single-arm study of 35 adult RRP patients, 51.4% (18/35) achieved a complete response, defined as requiring no surgical intervention in the 12 months following treatment. Among those responders, 83% (15/18) maintained response for at least two years, indicating durable benefit. Relative to pretreatment rates, 86% of patients had fewer surgeries in Year 1, 91% in Year 2, and 95% in Year 3; the median duration of complete response has not yet been reached.

Key lesson

When natural history is well characterized and baseline burden is objective, self-controlled comparisons can carry real evidentiary weight.

Case Study 2: Waskyra (Wiskott-Aldrich Syndrome Gene Therapy)

Design

Two single-arm multinational studies supplemented by an expanded-access program.

Population context

Only 27 patients were treated across all sources. Randomization was not feasible.

Primary clinical signal

A 93% reduction in severe infections comparing the 6–18 months post-treatment period with the 12 months prior to treatment.

Regulatory flexibility in practice

FDA exercised flexibility not by lowering standards, but by evaluating totality of evidence. Two independent studies showed consistent effects; expanded-access data reinforced durability; mechanistic data demonstrated genetic correction aligned with clinical benefit.

Key lesson

Single-arm trials rarely stand alone. Convergent evidence from multiple sources can substitute for randomization when those sources point in the same direction.

Zevaskyn: When Intra-Patient Randomization Replaces the Control Arm

Zevaskyn for recessive dystrophic epidermolysis bullosa (RDEB) used a particularly elegant and underutilized approach: randomized intra-patient control.

Patients with RDEB often have multiple chronic wounds. In the pivotal trial, each patient served as their own control: wounds were randomized to Zevaskyn or standard care, eliminating between-patient variability that would otherwise confound external comparisons.

The results were stark: 81% of treated wounds achieved ≥50% healing versus 16% of control wounds. FDA approved Zevaskyn largely on the basis of this single study because the design itself neutralized the usual objections to single-arm evidence.

Other recent approvals reinforce the same pattern:

• Kebilidi (AADC deficiency, November 2024): severe neurologic disease with clear functional endpoints and strong mechanistic rationale
• Ryoncil (SR-aGvHD, December 2024): life-threatening condition, no adequate alternatives, clinically meaningful outcomes
• Gomekli (NF1, February 2025): objective tumor-related endpoints in a genetically defined population

Different designs. Same underlying logic.

What Made These Single-Arm Programs Work

Clear Natural History.

Single-arm designs fail when variability overwhelms signal. These programs succeeded because the untreated disease course was already well understood: progressive trajectories, quantifiable baselines (infection rates, surgical frequency, functional decline), and outcomes that could be interpreted without a concurrent control. When you can credibly say, “this patient would have required five surgeries this year without treatment,” the counterfactual is no longer speculative, and a single arm becomes defensible.

Confirmatory Evidence Diversity.

None of these approvals rested on a single evidentiary pillar. Waskyra combined two independent studies with expanded-access data. Papzimeos paired clinical outcomes with mechanistic confirmation of HPV-specific T-cell responses. Zevaskyn was supported by eight years of Phase 1/2a follow-up demonstrating durability. The pattern is consistent: multiple independent lines of evidence converging on the same conclusion can substitute, in part, for the certainty randomization would have provided.

Objective, Interpretable Endpoints.

These programs measured direct clinical outcomes tied to patient burden: surgeries avoided, infections prevented, wounds healed. Not biomarkers. Not symptom scores. Not surrogate endpoints that require inferential leaps. When the endpoint is “the patient no longer needs surgery,” interpretation does not depend on modeling assumptions or comparator selection—it is self-evident.

Serious Disease and Unmet Need.

Regulatory flexibility is not evenly distributed. Every program here addressed life-threatening or severely disabling conditions with no adequate existing therapies, often driven by well-defined genetic mutations. FDA’s tolerance for single-arm evidence scales with both disease severity and the absence of alternatives. A single-arm trial in a mild disease with available treatments would face a very different reception.

Designing Your Single-Arm Trial: A Practical Framework

Before designing anything, answer four questions: Is the natural history well characterized? Do registries or historical cohorts exist to anchor comparisons? What measures demonstrate target engagement? Which outcomes reflect real patient benefit rather than convenient measurement?

If you can't answer confidently, a single-arm design is premature. You're not ready to argue that randomization is unnecessary; you're hoping FDA won't notice you skipped it.

Plan confirmatory evidence early. A single pivotal trial is rarely enough. The goal is convergence: nonclinical data, pharmacodynamic links, expanded-access experience, natural-history registries, all pointing the same direction. When sources conflict or remain sparse, the absence of a control arm stops being tolerable.

Statistical planning differs here. You're estimating event rates under natural history, defining clinically unambiguous effect sizes, and accounting for scrutiny that concurrent controls would have absorbed. When robust historical data exist, Bayesian borrowing can reduce enrollment, but only when pre-specified. Our Bayesian Borrowing calculator quantifies effective sample size while accounting for population drift.

Finally: request an RDDP or Type B meeting, but show up with a total-evidence argument, not a request for permission.

Why These Approvals Don’t Require a “Plausible Mechanism Pathway”

In November 2025, FDA leadership proposed a “plausible mechanism pathway” allowing approval based primarily on mechanistic rationale. Critics argued it lowered standards; proponents claimed it formalized existing practice for true n-of-1 therapies.

The 2025 rare-disease approvals suggest the debate is largely moot.

Waskyra showed that approval for 27 patients is possible under existing frameworks when evidence is structured correctly. Papzimeos demonstrated that self-controlled designs can support approval when disease course is well understood. Plausible-mechanism approaches work when mechanism and outcome are tightly coupled; they become problematic when that link is distant.

FDA’s existing flexibility, paired with rigorous evidence packages, already accommodates what sponsors need. The challenge is not new pathways—it is building credible evidence-totality arguments.

Conclusion

For biostatisticians, the challenge is no longer just sample-size calculation. It is evidence synthesis across heterogeneous sources: trials, natural-history registries, nonclinical models, mechanistic biomarkers, and expanded-access experience.

That is a different statistical mindset than planning a 1:1 randomized trial; and it is increasingly what rare-disease development demands.

Designing single-arm trials requires explicit modeling of natural-history uncertainty, effect size, delayed treatment effects, and regulatory scrutiny. Bayesian borrowing can reduce required sample sizes, but only when pre-specified and justified.

These are not optional refinements. They are fundamental to building single-arm trials that withstand regulatory review, and to earning the flexibility FDA is already prepared to grant.

📬 For more essays on experimental design, regulatory evidence, and statistical decision-making across domains, subscribe to the Evidence in the Wild newsletter or browse the archive.